Thank you for conveying these questions to us. We believe that questions are the primary driving force of scientific progress. Two concerns are related to possible bias in our study design because of differences in numbers between the maxillary and mandibular retainers and between the bonding surfaces of the test and control groups.
In our randomized, single-blind study, we randomly selected patients for the test and control groups to bond 2, 4, or 6 teeth according to the best benefit for each patient by the best possible clinical judgments. In small-scale clinical randomized studies, uneven numbers between the test and control groups in terms of sex, number of bonded surfaces, maxillary vs mandibular retainers, and so on are possible. In a larger randomized study, these tend to level out. During the study, we were aware of these issues as described. For evaluating our control group, we resorted to comparisons with reported results from larger studies: ie, references 15 and 16 in our article. The outcome of such comparisons showed that our control group’s data were similar to those of other studies with considerably larger patient groups. This proved that our control group was randomized enough. Since the test group was also randomized in the same way, we can trust the results of the test group. If the uneven numbers are purely due to random error, statistical procedures will handle the uneven numbers well.
Alternatively, a stratification study design might overcome the problems described. It requires a considerably larger patient group so that factors such as maxillary and mandibular retainers, number of teeth bonded, tooth type bonded, degree of overbite, dietary roughness, type of malocclusion (maxillary retainer at risk for Class II and vice versa in Class III), patients’ ethnic origins, bonding material used, operator factors, oral hygiene equipment used, and so on, can be considered. However, until it can be shown that bonding lingual retainers without liquid resin is a viable technique, it does not seem reasonable to subject large numbers of patients to a mode of treatment that might not give satisfactory results. Furthermore, it is not easy to ensure that all factors of interest have been identified and included in the sampling and stratification of the study. It is also not easy to ensure consistency of the factor studied; eg, patients in a soft-diet group might not have a soft diet at every meal during the study. Patient dropout is most frustrating in a carefully stratified study.
Stratification vs randomization could drag on for a few days of discussion or take a few book chapters to be explained perfectly. We think at different stages of a research project different sorts of techniques should be used. In our road map seeking for some answers to our questions about enamel adhesion and biocompatibility of dental composites, randomization is the technique of choice at this stage. If required, the stratification technique can be used in future studies on this subject.